Performance of a multi-channel adaptive Kalman algorithm for active noise control of non-stationary sources

Commonly used adaptive algorithms which determine the coefficients of a finite impulse response feed-forward filter in an active noise control application, as the filtered reference least mean squares algorithm, are not performing well when the sound source is non-stationary. A multiple input and multiple output Kalman algorithm potentially has a much better tracking performance, but has a few disadvantages, such as a high calculation complexity and the potential build-up of round-off errors. To overcome these problems a multi-channel Kalman algorithm is presented in fast-array form. The performance of this algorithm was tested in simulations and gives promising results for non-stationary sources, as long as the velocity of the sound source is relatively low. When the sound source has a higher velocity, the Doppler effect plays a significant role on the sound waves, giving a reduction in performance of the algorithm. Also the performance of the Kalman algorithm was validated in a real-time experiment. When the state space equations are rewritten, so the estimated impulse response of the secondary path is equal to a finite impulse response filter, the algorithm shows a comparable rate of convergence in experiments and simulations.